# US 2022/0036498 A1 · Methods and apparatus for mapping source location for input data to a graphics processing unit

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2022/0036498 A1
- Document ID: US20220036498A1
- Status: Published application
- Kind: A1
- Application number: 16/984,024
- Filed: 2020-08-03
- Published: 2022-02-03
- Earliest claimed: 2020-08-03
- Assignees: QUALCOMM Incorporated
- Applicants: QUALCOMM Incorporated
- Inventors: Liang LI; Elina KAMENETSKAYA; Andrew Evan GRUBER
- Web page: https://patentgrep.com/patent/US20220036498A1

## CPC

- G06T 1/20: Processor architectures; Processor configuration, e.g. pipelining
- G06F 9/3001: {Arithmetic instructions}
- G06F 9/30032: {Movement instructions, e.g. MOVE, SHIFT, ROTATE, SHUFFLE}
- G06F 9/30036: {Instructions to perform operations on packed data, e.g. vector, tile or matrix operations}
- G06F 9/30145: {Instruction analysis, e.g. decoding, instruction word fields}
- G06F 9/3016: {Decoding the operand specifier, e.g. specifier format}
- G06F 9/5066: {Algorithms for mapping a plurality of inter-dependent sub-tasks onto a plurality of physical CPUs (mappping at compile time, see G06F8/451)}
- G06F 17/142: {Fast Fourier transforms, e.g. using a Cooley-Tukey type algorithm}
- G06F 17/16: Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)}
- G06T 1/60: Memory management
- G06T 15/005: {General purpose rendering architectures}
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

The present disclosure relates to methods and apparatus for mapping a source location of input data for processing by a graphics processing unit. The apparatus can configure a processing element of the graphics processing unit with a predefined rule for decoding a data source parameter for executing a task by the graphics processing unit. Moreover, the apparatus can store the parameter in local storage of the processing element and configure the processing element to decode the parameter according to the at least one predefined rule to determine a source location of the input data and at least one relationship between invocations of the task. The apparatus can also load, to the local storage of the processing element, the input data from a plurality of memory addresses of the source location determined by the parameter. A one logic unit can then execute the task on the loaded input data.

## Claims

### Claim 1

A method for mapping a source location of input data for processing by a graphics processing unit, the method comprising:
configuring a processing element of the graphics processing unit with at least one predefined rule for decoding a data source parameter for a task to be executed by the graphics processing unit;
storing the data source parameter in local storage of the processing element;
decoding, by the processing element, the data source parameter according to the at least one predefined rule to determine a source location of the input data and at least one relationship between invocations of the task;
loading, to the local storage of the processing element, the input data from a plurality of memory addresses of the source location that are determined by the decoded data source parameter; and
executing the task, by at least one logic unit of the processing element, on the loaded input data to generate output data.

### Claim 2 (depends on 1)

The method of claim 1, further comprising configuring the at least one logic unit to generate the output data based on a matrix multiply operation.

### Claim 3 (depends on 2)

The method of claim 2, wherein the input data comprises at least one input matrix and the data source parameter comprises a single element value within the at least one input matrix.

### Claim 4 (depends on 3)

The method of claim 3, further comprising configuring the at least one logic unit to decode the single element value to determine the source location of the plurality of memory addresses.

### Claim 5 (depends on 4)

The method of claim 4, wherein each data input value of the at least one matrix is stored in a respective memory address of the plurality of memory addresses.

### Claim 6 (depends on 3)

The method of claim 3, further comprising calculating, by the at least one logic unit, the output data by multiplying the at least one input matrix by one or more additional input matrices to generate the output data.

### Claim 7 (depends on 1)

The method of claim 1, further comprising storing the output data in the local storage of the processing element.

### Claim 8 (depends on 1)

The method of claim 1, further comprising defining a mode of operation for the at least one logic unit to execute the task.

### Claim 9 (depends on 8)

The method of claim 8, wherein the task comprises a fast Fourier transform operation of the loaded input data to generate the output data.

### Claim 10 (depends on 9)

The method of claim 9, further comprising configuring the at least one logic unit to determine shuffle instructions for the fast Fourier transform operation based on the defined mode of operation.

### Claim 11 (depends on 1)

The method of claim 1, wherein the task executed by the at least one logic unit is a deep learning operation.

### Claim 12 (depends on 11)

The method of claim 11, wherein the input data is image data and the output data for the executed deep learning operation indicates at least one of an identification of content of the image data, a prediction of a next image in an image sequence of the content including the image data, and an identification of a missing part of the image data.

### Claim 13 (depends on 1)

The method of claim 1, further comprising loading the input data to the local storage of the processing element from the electronic memory that is located external to the graphics processing unit.

### Claim 14 (depends on 1)

The apparatus of claim 1, wherein the processing element comprises a shader unit of the graphics processing unit.

### Claim 15

An apparatus for mapping a source location of input data for processing by a graphics processing unit, the apparatus comprising:
a memory; and
at least one processor coupled to the memory and configured to:
configure a processing element of the graphics processing unit with at least one predefined rule for decoding a data source parameter for a task to be executed by the graphics processing unit;
store the data source parameter in local storage of the processing element;
configure the processing element to decode the data source parameter according to the at least one predefined rule to determine a source location of the input data and at least one relationship between invocations of the task;
load, to the local storage of the processing element, the input data from a plurality of memory addresses of the source location that are determined by the decoded data source parameter; and
configure at least one logic unit of the processing element to execute the task on the loaded input data to generate output data.

### Claim 16 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is further configured to configure the at least one logic unit to generate the output data based on a matrix multiply operation.

### Claim 17 (depends on 16)

The apparatus of claim 16, wherein the input data comprises at least one input matrix and the data source parameter comprises a single element value within the at least one input matrix.

### Claim 18 (depends on 17)

The apparatus of claim 17, wherein the at least one processor is further configured to configure the at least one logic unit to decode the single element value to determine the source location of the plurality of memory addresses.

### Claim 19 (depends on 18)

The apparatus of claim 18, wherein each data input value of the at least one matrix is stored in a respective memory address of the plurality of memory addresses.

### Claim 20 (depends on 17)

The apparatus of claim 17, wherein the at least one processor is further configured to control the at least one logic unit to calculate the output data by multiplying the at least one input matrix by one or more additional input matrices to generate the output data.

### Claim 21 (depends on 15)

The apparatus of claim 15, wherein the at least one processor is further configured to store the output data in the local storage of the processing element.

### Claim 22 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is further configured to define a mode of operation for the at least one logic unit to execute the task.

### Claim 23 (depends on 22)

The apparatus of claim 22, wherein the task comprises a fast Fourier transform operation of the loaded input data to generate the output data.

### Claim 24 (depends on 23)

The apparatus of claim 23, wherein the at least one processor is further configured to configure the at least one logic unit to determine shuffle instructions for the fast Fourier transform operation based on the defined mode of operation.

### Claim 25 (depends on 15)

The apparatus of claim 15, wherein the operation executed by the at least one logic unit is a deep learning operation.

### Claim 26 (depends on 25)

The apparatus of claim 25, wherein the input data is image data and the output data for the executed deep learning operation indicates at least one of an identification of content of the image data, a prediction of a next image in an image sequence of the content including the image data, and an identification of a missing part of the image data.

### Claim 27 (depends on 15)

The apparatus of claim 15, wherein the at least one processor is further configured to load the input data to the local storage of the processing element from the electronic memory that is located external to the graphics processing unit.

### Claim 28 (depends on 15)

The apparatus of claim 15, wherein the processing element comprises a shader unit of the graphics processing unit.

### Claim 29

An apparatus for mapping a source location of input data for processing by a graphics processing unit, the apparatus comprising:
means for configuring a processing element of the graphics processing unit with at least one predefined rule for decoding a data source parameter for a task to be executed by the graphics processing unit;
means for storing the data source parameter in local storage of the processing element;
means for configuring the processing element to decode the data source parameter according to the at least one predefined rule to determine a source location of the input data and at least one relationship between invocations of the task;
means for loading, to the local storage of the processing element, the input data from a plurality of memory addresses of the source location that are determined by the decoded data source parameter; and
means for controlling at least one logic unit of the processing element to execute the task on the loaded input data to generate output data.

### Claim 30

A computer-readable medium storing computer executable code for mapping a source location of input data for processing by a graphics processing unit, the code when executed by a processor causes the processor to:
configure a processing element of the graphics processing unit with at least one predefined rule for decoding a data source parameter for a task to be executed by the graphics processing unit;
store the data source parameter in local storage of the processing element;
configure the processing element to decode the data source parameter according to the at least one predefined rule to determine a source location of the input data and at least one relationship between invocations of the task;
loading, to the local storage of the processing element, the input data from a plurality of memory addresses of the source location that are determined by the decoded data source parameter; and
control at least one logic unit of the processing element to execute the task on the loaded input data to generate output data.
